Table 1
Detailed breakdown of feature dimensions and model parameters by component.
| Component/modality | Extracted dim | Time-freq used | Unified dim | Param dount (k) |
|---|---|---|---|---|
| Feature extraction | ||||
| Time-domain features (current, displacement) | 12 | No | — | — |
| Time–frequency features (vibration, acoustic) | 24 | Yes | — | — |
| Statistical features (all modalities) | 24 | No | — | — |
| Graph neural network (GCN) | ||||
| Input layer (30 → 64) | — | — | — | 2.0 |
| Hidden layer 1 (64 → 32) | — | — | — | 2.1 |
| Hidden layer 2 (32 → 16) | — | — | — | 0.5 |
| Graph embedding output | — | — | 16 | 30.4 (Subtotal) |
| Fusion and Classification | ||||
| Modality-specific FC layers (5×) | — | — | 64 | 42 + 78 + 78 + 42 + 35 = 275 |
| Channel attention FC layers | — | — | 8→5 | 0.04 |
| Spatial attention conv layer | — | — | 32 | 0.1 |
| Classifier FC layers (272→128→64→13) | — | — | — | 35.0 |
| Total model parameters | 275 | |||
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.